In kinship testing, the level of genotypic similarity between individuals at suitable genetic markers is utilised to assess their degree of familial relationship. Inference is made either by evaluating the available genotypic information under different hypotheses about kinship, represented by different pedigree structures, or through the direct quantification of kinship coefficients using a statistical approach. One complicating factor in genetic kinship testing, however, is population structure whereby the assumption of statistical independence between genotypes of unrelated individuals, or between genotypes at unlinked markers, may become violated. Similarly, physical linkage between markers can also render the simple, so-called ‘product rule’ for composite likelihood calculation invalid. In these instances, specific statistical approaches and computer algorithms need to be used that properly adjust the relevant computations for population structure or linkage.

Key Concepts:

• Kinship testing involves evaluation of the level of genotypic similarity between individuals at selected genetic markers.
  
• The optimal basis for decision-making in kinship testing is the likelihood ratio of different hypotheses about kinship, taking the available genetic evidence into account.
  
• Posterior odds for given hypotheses about kinship can be calculated from prior odds and likelihood ratios, using Bayes formula.
  
• Coancestry as arising, for example, from population structuring renders the calculation of multimarker likelihoods by simple multiplication of single-marker likelihoods (known as the ‘product rule’) invalid.
  
• General kinship testing is equivalent to likelihood calculations of genotypes in pedigrees, which in turn must take physical linkage between markers into account.
  
• The degree of kinship between two individuals is quantified by their coefficient of relatedness, which equals the expected number of alleles shared identical-by-descent from a common ancestor.
  
• The coefficient of relatedness is either calculated for a given type of relationship, and then used to predict genetic similarity, or estimated from observed levels of genetic similarity to infer pair-wise relatedness.